Skip to content

GitLab

Commit 5a9db327 authored by zhuwenwen's avatar zhuwenwen
Browse files

update fa interface param

parent 50ed6d0a
Show whitespace changes
Inline Side-by-side
Showing with 4 additions and 4 deletions
vllm/attention/backends/rocm_flash_attn.py
View file @ 5a9db327
...@@ -449,8 +449,8 @@ class ROCmFlashAttentionImpl(AttentionImpl): ...@@ -449,8 +449,8 @@ class ROCmFlashAttentionImpl(AttentionImpl):
v=value, v=value,
cu_seqlens_q=prefill_meta.seq_start_loc, cu_seqlens_q=prefill_meta.seq_start_loc,
cu_seqlens_k=prefill_meta.seq_start_loc, cu_seqlens_k=prefill_meta.seq_start_loc,
max_seqlens_q=prefill_meta.max_prefill_seq_len, max_seqlen_q=prefill_meta.max_prefill_seq_len,
max_seqlens_k=prefill_meta.max_prefill_seq_len, max_seqlen_k=prefill_meta.max_prefill_seq_len,
softmax_scale=self.scale, softmax_scale=self.scale,
causal=True, causal=True,
) )
...@@ -526,8 +526,8 @@ class ROCmFlashAttentionImpl(AttentionImpl): ...@@ -526,8 +526,8 @@ class ROCmFlashAttentionImpl(AttentionImpl):
v=value, v=value,
cu_seqlens_q=prefill_meta.seq_start_loc, cu_seqlens_q=prefill_meta.seq_start_loc,
cu_seqlens_k=prefill_meta.seq_start_loc, cu_seqlens_k=prefill_meta.seq_start_loc,
max_seqlens_q=prefill_meta.max_prefill_seq_len, max_seqlen_q=prefill_meta.max_prefill_seq_len,
max_seqlens_k=prefill_meta.max_prefill_seq_len, max_seqlen_k=prefill_meta.max_prefill_seq_len,
softmax_scale=self.scale, softmax_scale=self.scale,
causal=True, causal=True,
) )
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment